Density plots
## png
## 2
## png
## 2
## png
## 2
Deconvolve the control distribution in 2 gaussians using GMM
First decide which gaussians need to be separated
## [1] "2080"
## number of iterations= 352
## [1] "5050"
## number of iterations= 323
## [1] "8020"
## number of iterations= 326
## [1] "BT112"
## number of iterations= 730
## [1] "BT245"
## number of iterations= 121
## [1] "BT248"
## number of iterations= 204
## [1] "BT320"
## number of iterations= 387
## [1] "BT333"
## number of iterations= 668
## [1] "BT360"
## number of iterations= 537
## [1] "BT569"
## WARNING! NOT CONVERGENT!
## number of iterations= 1000
## [1] "LBT003"
## number of iterations= 213
## [1] "LBT005"
## WARNING! NOT CONVERGENT!
## number of iterations= 1000
## [1] "LBT059"
## number of iterations= 165
## [1] "LBT062"
## number of iterations= 105
## [1] "LBT070"
## number of iterations= 161
## [1] "LBT086"
## WARNING! NOT CONVERGENT!
## number of iterations= 1000
## [1] "LBT124"
## number of iterations= 378
For every cell/marker, we calculate the probability of belonging to the control distribution. Two tails: P(value>red) P(value<red)
The distribution of p-values looks more or less uniform -> good QC.
We define as induction for a marker those cells with a p-value < 0.025 for either tail. p-low < 0.025 (-1), p.high < 0.025 (+1), else 0.
## png
## 2
MDM2, p21, p53
## png
## 2
## png
## 2
Fit all panels for 1, 2, …, n markers and build a pareto front based on performance ()
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
The flow will be: First, filter based on MDM2 samples Second, regression based on p21 model
Since p21 is only one marker we can just take the % of induced cells as predictor
## png
## 2
Select panel of size 1 as the best performing one. Both BAX and p21 have a comparable predictive ability. Choose p21 since the effect size (shift in the distribution) was larger. This would make a more reliable assay. In BAX the differences are too small (yet significant).
For each sample, calculate the relative proportion of cells in each fingerprint and do DR on it.
Project biopsies
## png
## 2
## png
## 2
## png
## 2
pH2AX, p21, p53
## png
## 2
## png
## 2
Fit all panels for 1, 2, …, n markers and build a pareto front based on performance ()
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
## png
## 2
For each sample, calculate the relative proportion of cells in each fingerprint and do DR on it.
## class: SlingshotDataSet
##
## Samples Dimensions
## 17 2
##
## lineages: 1
## Lineage1: RES MIXED SENS
##
## curves: 0
## png
## 2
## png
## 2
## png
## 2
##
## Call:
## lm(formula = tmp_samples$AUC ~ tmp_samples$pseudotime)
##
## Residuals:
## Min 1Q Median 3Q Max
## -20.3185 -7.2990 -0.0163 7.0258 24.5729
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 77.496 4.889 15.851 8.88e-11 ***
## tmp_samples$pseudotime -27.157 5.449 -4.984 0.000163 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 11.32 on 15 degrees of freedom
## Multiple R-squared: 0.6235, Adjusted R-squared: 0.5984
## F-statistic: 24.84 on 1 and 15 DF, p-value: 0.0001633
## png
## 2
##
## Call:
## lm(formula = tmp_samples$MaxInhibition ~ tmp_samples$pseudotime)
##
## Residuals:
## Min 1Q Median 3Q Max
## -37.970 -6.847 0.929 8.955 22.845
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 72.426 6.461 11.210 1.09e-08 ***
## tmp_samples$pseudotime -37.381 7.201 -5.191 0.00011 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 14.95 on 15 degrees of freedom
## Multiple R-squared: 0.6424, Adjusted R-squared: 0.6186
## F-statistic: 26.95 on 1 and 15 DF, p-value: 0.0001095
## png
## 2
## png
## 2
Project biopsies
## png
## 2
## png
## 2